1. Field of the Invention
The present invention relates to a device for detecting an amount of skew in a document sheet.
2. Description of the Related Art
Image processing apparatuses for apply processing after reading an image depicted on a document have been provided with an automatic document feeder. Examples of such an image processing apparatus include a copier and a multi-functional peripheral (MFP: multifunction device or combination machine). The automatic document feeder is operable to automatically feed document sheets placed thereon, one by one, to a scan position.
Depending on what a document is like or how the document is placed on the automatic document feeder, the document fed by the automatic document feeder is sometimes skewed (tilted). An example of well-known methods for correcting skew in a document is a method of bringing the document into contact with a registration roller to warp the document. Unfortunately, the method involves making a noise between the registration roller and the document, which degrades the low-noise running.
In view of this, another technology has recently received attention. According to the technology, skew itself in a document is not corrected. To be specific, image processing is performed of scanning an image recorded on the skewed document and rotating the scanned image, so that an image having no skew is generated.
Before skew correction by the image processing, it is necessary to detect an amount of skew in a document. Conventional technologies for detection of an amount of skew in a document include the technologies described in Japanese Laid-open Patent Publication Nos. 2003-259089, 2003-259090, 2003-319160, 2009-042814, 2002-199179, 2013-146034, and 2015-015687.
The technology described in Japanese Laid-open Patent Publication No. 2003-259089 discloses the following: when an original image is read, a shadow due to the thickness of the original is also read at once. The inclination angle of an end of the original is detected based on the positions of black pixels due to the shadow. The technology described in Japanese Laid-open Patent Publication No. 2003-259090 discloses applying density conversion to sharpen an edge of a shadow in an image read out. The technology described in Japanese Laid-open Patent Publication No. 2003-319160 discloses sharpening a shade part by filtering using a Laplacian filter.
Using Hough transform to determine a side of an image is disclosed in Japanese Laid-open Patent Publication Nos. 2009-042814, 2002-199179, 2013-146034, and 2015-015687. The determination of the side tells an inclination of the side. The technology described in Japanese Laid-open Patent Publication No. 2015-015687 discloses using a method of least squares as a method for detecting a skew angle.
An amount of skew in a document can be detected more precisely by performing image processing for the detection than by using a sensor to detect whether or not a document is present. The detection without such a sensor leads to reduction in cost of components.
However, the detection of an amount of skew in a document by image processing involves a problem. The problem is how an amount of skew is detected rapidly and accurately based on gradation values of pixels of an image.
In detecting an amount of tilt of a shadow of the original image as described in Japanese Laid-open Patent Publication Nos. 2003-259089, 2003-259090, and 2003-319160, a tilt amount thereof cannot be detected, or, even if the detection is made, an error in the tilt amount is sometimes large in the following cases: where a shadow has a large width due to thick original documents or upward warp of the entire edge of the original; and where a shadow has an uneven width due to upward warp of a partial edge of the original. In short, such conventional technologies involve a problem that a state of the original influences accuracy in the detection.
Determining a side of an image by Hough transform as disclosed in Japanese Laid-open Patent Publication Nos. 2009-042814, 2002-199179, 2013-146034, and 2015-015687 involves a problem of taking a long time to perform image processing for skew detection.